Amelia's Weekly Fish Fry

Adventures in Memory - Samsung’s HBM-PIM and A New Brain-Inspired Memory Device

In this week’s Fish Fry podcast, Nam Sung Kim (Senior Vice President of Samsung’s Memory Business Unit) joins me to discuss why he believes our computing architectures must evolve to meet the needs of artificial intelligence and machine learning applications and how breaking the traditional Von Neumann processor-memory boundary could be disruptive to both the hardware and software sides of our industry, and the details of Samsung’s new HBM Processing-In Memory. Also this week, I investigate a new brain-inspired memory device developed by the University of Singapore and why this new novel molecular memristor could represent a significant breakthrough in our quest to design low-energy computing. 

Crafting the Next Generation of AI - Deci is Looking to Change How We Build Artificial Intelligence

It’s Friday and we’re taking a deep dive into the world of artificial intelligence! Yonatan Geifman (Co-founder and CEO - Deci) joins me this week to discuss how we can use AI itself to craft the next generation of AI. We also chat about how developers can streamline artificial intelligence development and where AI is headed in the next couple years and decades to come. Also this week, I take a closer look at new research from an international group of researchers that aims to answer the question: What if photonics can help us better recognize patterns for machine learning?

 

Machine Learning and The New Age of Intelligent Chip Design

Machine learning is once again taking center stage in this week’s Fish Fry podcast! Rod Metcalfe (Cadence Design Systems) joins me to discuss the role of machine learning in the future of chip design, why implementation teams are having a hard time keeping up with today’s advanced node designs, and how advancements in computer science are helping usher in a new age of intelligent chip design. I also take a closer look at how a team of researchers is using machine learning to help us more accurately predict how new materials deform, fail and even transfer heat at the atomic scale.

Better than a Human Driver: Trusted Electronics and Functional Safety

In today’s Fish Fry podcast, we have a virtual sundae of electronic engineering goodness. We've got a scoop of autonomous driving, a scoop of IIoT, a little Infineon secret sauce, and a whole lot of sprinkles… because what good is a sundae without sprinkles, right? First up, Phil Hutchinson (element14) joins me to discuss element14’s “Low Power IoT Design Challenge with Infineon Technology”. We take a closer look at the details of this contest including the motivations behind its creation and why collaboration is a crucial component with this design challenge. Next up, Bill Stewart (Infineon) joins me to chat about the role trusted electronics play in the realm of automotive design and what he thinks it will take to make autonomous vehicle technology more widespread. 

Machine Vision Meets Hackathon: ADLINK’s 20/20 Hack Solves Real World Problems

Who loves a good ol’ fashioned hackathon? In this week’s Fish Fry podcast, we are talking about encouraging innovation in machine learning with ADLINK’s recent 20/20 Vision Hack. Paul Wealls (ADLINK) and Sergio Velmay join me to discuss the motivations behind creation of this hackathon, the details of Sergios’s winning project entry, and how ADLINK and Intel are looking to help solve  real-world problems in manufacturing, logistics and industrial applications with this unique program. Also this week, I check out how a group of engineers from the University of Illinois (with the help of two supercomputers) have developed a way to use physics-informed neural networks to predict the outcomes of the complex processes involved in additive manufacturing.

 

Let's Get Speedy! Digital Signal Processing for Always-On Systems, Sensor Fusion and Computer Vision

What would a Friday be without a deep dive into digital signal processing for computer vision and AI? In this week’s Fish Fry podcast, Pulin Desai (Cadence Design Systems) joins me to discuss the need for increased speed in computer vision algorithms today, the unique processing requirements for sensor fusion, 3D capture, always-on systems, and the details of Cadence’s new Vision Q8 and Vision P1 DSPs. Also this week, I take a closer look at a new research study at the University of Gothenburg that contends that machine learning may be the key to finding effective testing methods during epidemic outbreaks.

 

The Digital Thread: The Advancement of Digital Backbones and the World’s First Digital Fiber Fabric

Ready, Aim, Digitize! In this week’s podcast, we start things off by taking a closer look at the world’s first digital fiber fabric developed at MIT. We investigate how this new digital fiber is able to collect and store data and the role AI plays in how it is able to process that data. Also this week, Duc Huy Tran (Aitech) joins us to discuss the enablement of digital backbones for military applications, the evolution of COTs in the military and aerospace designs, and why standardization is crucial in this arena.

 

Computations in Memory: Imec’s Analog Inference Accelerator Brings Extreme Energy Efficiency to Neural Network Acceleration

The internet of things takes center stage in this week's Fish Fry podcast! Peter De Backer (imec) joins me to discuss the challenges of developing neural networks for IoT devices and the details of imec’s Analog Inference Accelerator (AnIA).  Also this week, I take a closer look at how a team of researchers from National University of Singapore (NUS) and Japan's Tohoku University (TU) used spin-torque oscillators to harvest and convert wireless radio frequencies into energy.  

 

The World According to Analog: Why Our Digital Futures Depend on Analog

In this week’s Fish Fry podcast, we investigate the analog data conundrum with Tom Doyle and Marcie Weinstein from Aspinity. We take a closer look at the power and data challenges of processing at the edge, what Aspinity’s new analog machine learning IC is all about, and the details of their new evaluation kits that utilize their analogML core. Also this week, I dig into the details of a new organic, metal free, non-toxic battery developed by Texas A&M University. (Spoiler Alert: It can be recycled by breaking it down in acid!)

Research meets Realization: InnovationLab Brings Organic Printed Electronics into the Mainstream

I’ve got three words for you: Printed Organic Electronics! My distinguished guest is Dr. Florian Ullrich from InnovationLab. Florian and I are talking all about the unique printing technology behind InnovationLab’s printed electronics, the details of InnovationLab’s collaboration with Evonik to produce the first fully printed rechargeable batteries for flexible printed sensors, and how InnovationLab’s shareholder structure encourages innovation big and small.